An integrated approach to exploring soil fertility from the perspective of rice (Oryza sativa L.) yields

Abstract

Soil fertility should be explored from a productivity perspective because the most important function of the soil is to ensure crop yield. This study presents an integrated and scientific approach to exploring soil fertility based on rice (Oryza sativa L.) yields, using Jinxian County as an example. The main soil types are Haplic Acrisols and Hydragric Anthrosols. Five soil fertility indicators (p-Value < 0.05) were selected according to the generalized additive model (GAM) hypothesis test between the rice yield and each indicator. Furthermore, these indicators were used to assess the quality of soil fertility via Takagli and Sugeno (T-S) fuzzy neural networks models. Finally, the geodetector model was used to explore the restricting indicators of soil quality that can influence rice yields. The results indicate that continuous fertilization for decades has improved the surface soil organic matter (SOM) concentrations in the study area. However, the surface total potassium (KT) was still low at a mean value of 13.95 ± 4.74 mg/kg. The determination power (DP) of exchangeable magnesium (0–20 cm) and KT (20–40 cm) of 0.067 and 0.061, respectively, revealed that the proper use of potassium and magnesium fertilizers can increase rice yields. This integrated soil fertility exploration could scientifically assess the soil fertility and identify mainly restricting indicators of soil fertility in the study area. Moreover, these effective models with minor adjustments could be applied to assess soil fertility in other typical areas.